Predicting the Unseen: Integrating machine-learning with novel detection methods to reveal hidden malaria transmission reservoirs

Malaria is caused by Plasmodium parasites and transmitted between humans by mosquitoes. Despite being a preventable and treatable disease, malaria was estimated by the 2025 WHO World Malaria Report to have caused 82 million cases and 610,000 deaths globally in 2024. Of these, the WHO African region is disproportionately burdened, accounting for up to 95% of reported cases. Although malaria control interventions led to a marked decline in disease incidence across the continent from 2000 to 2015, a resurgence in morbidity and mortality has since been reported, with an 8% increase observed in 2023. Several factors, including the spread of drug resistance, changes in mosquito vector behaviour, and poor implementation of preventive control interventions, compound this disease resurgence. However, climate change is proving to be the main accelerator.

By forecasting both the timing and severity of outbreaks, this framework enables the precise deployment of community-level prevention and treatment strategies where they are needed most.

Seasonal shifts, changes in rainfall patterns, and warmer temperatures, together with human behavioural responses, directly contribute to malaria outbreaks. While predictive mathematical models that integrate human, climate, and vector data are essential for early warning systems, they are often limited by static datasets and an inability to account for non-linear shifts in transmission. To overcome these limitations, FAR-LeaF fellow Dr Jacob Agyekum is leveraging machine learning (ML) tools to develop a robust, integrated model to predict future outbreaks and severity indicators in Ghana. By leveraging MLs to process dynamic datasets, his project provides the computational framework to move beyond traditional modelling.

Apart from climatic data, predictive modelling requires accurate yet dynamic case surveillance data. In low-density regions, individuals carrying the transmissible stages of Plasmodium parasites are usually asymptomatic, either because they have completed treatment or have some form of immunity. Therefore, these individuals contribute to a hidden reservoir that sustains disease transmission. Technological and infrastructure constraints, combined with the lack of diagnostic tools capable of distinguishing between the pathogenic and transmissible stages of Plasmodium parasites, mean these transmissible reservoirs are not included in general surveillance strategies. Thus, current modelling systems are built on incomplete disease transmission dynamics. To address this, the project of a FAR-LeaF fellow, Dr Dina Coertzen, aims to develop a novel quantitative method to detect and distinguish between the pathogenic and transmissible stages of Plasmodium parasites. This method should be sufficiently robust to be used in low-density, resource-scarce endemic settings, such as those found in neighbouring regions in South Africa. With this work, she can identify hidden reservoirs sustaining malaria transmission. This level of data will provide a more comprehensive understanding of transmission dynamics.

Crucially, by integrating Coertzen's data, which identifies previously hidden transmissible reservoirs, into Agyekum's informed ML framework, they are reinventing how transmission dynamics are predicted. Therefore, this model can more accurately reflect true transmission dynamics and provide useful intelligence for community-based interventions. The integrated approach moves beyond localised snapshots to create a multiregional early warning system. By forecasting both the timing and severity of outbreaks, this framework enables the precise deployment of community-level prevention and treatment strategies where they are needed most.

Dina Coertzen, University of Pretoria, Institute for Sustainable Malaria Control, Department of Biochemistry, Genetics and Microbiology, Faculty of Natural and Agricultural Sciences, & Jacob Agyekum, Council for Scientific and Industrial Research, Water Research Institute, Ghana